Infants with subdural haematoma constitute a medical emergency, but they also immediately trigger thoughts of possible inflicted injury. Two articles in this issue highlight once again the incidence, aetiology, and neuroimaging of infantile subdural haematomas/effusions (SDH/E).

Hobbs et al report an incidence of subdural haematoma/effusion in infancy from all causes of 24.1 per 100 000 children less than 12 months of age (and 12.54 per 100 000 aged 0–2) in the largest UK study to date.¹ Cases of SDH/E diagnosed on brain imaging or at postmortem examination were reported to the BPSU, through the monthly reporting card system, over a 12 month period (April 1998–March 1999) from all specialists likely to have contact with infantile subdural haematoma/effusion. Because a subdural haematoma/effusion is a dynamic pathology and can be due to trauma or infection etc, the authors have very appropriately defined case entry as any child under 2 years with subdural haemorrhage, haematoma, or effusion, and have reported the aetiology that was ascribed by local reporter personnel.

A traumatic acutesubdural haemorrhage is a venous haemorrhage resulting from vascular shearing of cortical surface and interhemispheric bridging veins when motion of the brain within the skull stretches and tears the weakest of some 15–20 small bridging veins as they cross the subdural space. In the acute haemorrhage the brain is covered with a layer of undiluted blood clot with a high haematocrit. It is temporally considered acute if within three days of the injury.²

As the acute bleed breaks down by fibrinolysis, water is drawn into the collection, haemodilution occurs, and the haematocrit falls to less than 10%. There is subsequently a marked expansion after about one week. Denaturation of the haemoglobin changes the colour from bright red to brown and the collection, now 3 days to 3 weeks from the injury, is a subacute haemorrhage or subdural haematoma.

A chronic subdural collection is liquid with an even lower haematocrit—that is, more water than blood, and is more correctly termed a subdural effusion. In most cases this is more than three weeks from the injury and is a failure of the acute haematoma to resolve. It has two components: a liquid watery component, and fresh bleeding of a membrane. The membrane (which is vascular and bleeds easily) encapsulates the haematoma, and binds it to the dura where it undergoes degradation and invasion by fibroblasts. It may calcify after about three months. Incorporation of the haematoma into the dura as a membrane is the basis of the healing process. Labelled albumin can move freely in and out of the haematoma and bloodstream, and between the CSF and the subdural effusion. Infections cause subdural effusions largely by an exudative process, and these may also be associated with a membrane. Benign external hydrocephalus may simulate the density of a chronic subdural haematoma, and a FLAIR MRI is necessary to distinguish a proteinacious collection. It arises from an expanded subarachnoid space, and may predispose to bridging vein rupture following minor head injury.

The process by which an acute haemorrhage becomes chronic may depend on a critical volume of blood in the haemorrhage which defeats the repair process. The liberation of idiogenic osmoles by protein degradation, and the high protein content causes osmotic forces to convert the thin initial haemorrhage into a larger effusion. There are always fresh uncrenated cells in the fluid of even very chronic haematomas with persistent xanthochromia, suggesting there is continued fresh bleeding. This repeated bleeding may result from distraction of the brain from the inner table of the skull and so continued traction on the bridging veins, in addition to bleeding from the vascular membrane. There is therefore a pachymeningitis with a chronic inflammatory response seen in the meninges initiated by continued breakdown of erythrocytes. This inflammatory process causes vascular proliferation with the formation of sinusoids, fibrin deposition, fibrosis, and the formation of the membrane. Continued bleeding will maintain the inflammatory process and thicken the membrane in layers. Additionally the high fibrinolytic activity in the fluid may prevent the vessels sealing properly with a good clot. The presence of a membrane therefore confirms the chronic nature of the subdural collection, but an absence of one does not exclude a chronic collection.

A figure of 14.2 per 100 000 children less than 1 year (57% of all cases) reported by Hobbs et al for this UK and Republic of Ireland study, may represent the most accurate incidence for subdural haematoma/effusion due to non-accidental head injury (NAHI). Although regional differences in incidence will be minimised in a large study, ascertainment difficulties may be increased. Their figure is considerably less than that estimated from the Severnside Study,³ and approximately 40% less than in the Scottish study.⁴ However, as pointed out by the authors, case definition may not make them strictly comparable.

In more than one quarter of the cases in this study the diagnosis was delayed for more than a week after admission, emphasising that because infants with subdural haematoma/effusion present with a variety of non-specific clinical features (drowsy, poor feeding, irritability, fits, pallor, floppiness, etc) it is important to consider a haematoma/effusion as a possible cause in any unwell infant. Hobbs et al also noted that fewer diagnostic investigations were performed in the non-NAHI cases. This may be because other causes of the haematoma/effusion were obvious (for example, road traffic accident), or more easily diagnosed (for example, meningitis), or that the suspicion of an inflicted injury has influenced clinicians to undertake more detailed investigations to exclude other rarer causes of subdural collections, anticipating that their diagnosis of NAHI will later be scrutinised in child protection hearings and courts.

In the second paper on subdural haemorrhage in this issue, Datta et al, in a retrospective analysis of 74 cases of children under 2 years with a subdural haemorrhage who had neuroimaging, found that 49 (66%) were due to NAHI.⁵ The authors accepted a secure “non-accidental” diagnosis only in those cases confirmed by a perpetrator’s confession or conviction with supportive evidence of unexplained extracranial injuries. Their non-accidental cohort was similar in make up to other published series, in that it included 16 (33.2%) who had additional evidence of impact, seen as either a skull fracture, scalp swelling, or both.

The first neuroimaging in 90% of their cases was performed in a district hospital, and the authors emphasise that important decisions are often taken on the first brain scans. The first set of scans are, in our experience, often the only ones which permit an accurate interpretation of antecedent events. If subdural taps and aspirations are performed after the first scan, the subsequent density patterns of the subdural haematoma/effusion on later computed tomography (CT) scans are affected, leading to debate about whether this is new active bleeding, or whether the new subdural appearance represents layering within the haematoma from haematocrit changes, or whether it is due to new subdural bleeding from the needle puncture or arachnoidal tear, or resulting from new acute bleeding following excessive drainage of large amounts of surface subdural fluid. Subdurals on magnetic resonance imaging (MRI) can not at any time be dated, either on the first or subsequent scans. Datta et al therefore concur with Jaspan and colleagues⁶ that the quality, particularly of the first scan, should be standardised. The authors warn that suboptimal protocols, thick brain slices, and the absence of contrast enhancement, will increase the possibility of an inaccurate diagnosis.

The accuracy of the interpretation of a properly obtained CT or MRI scan is also paramount, particularly in child protection investigations. As Datta et al found, in 7/49 (14.3%) NAHI cases, the original CT interpretation was at first falsely reported as normal and not suggestive of NAHI, although this was later correctly interpreted after clinical suspicion prompted a radiological review. In a previous analysis of “missed cases” of child abuse, Jenny and colleagues⁷ reported that 31.2% (54/173 abused children with head injury) did not have a correct diagnosis recognised, and it took a mean of 7 days further to correct the diagnosis. In seven of her unrecognised cases (13%) there was misinterpretation of the radiological studies which contributed to the delay—a figure similar to that of Datta et al. As a result of delaying the correct diagnosis, Jenny reported 28% were re-injured and 40.7% experienced medical complications related to the misdiagnosis. Four of the five deaths in her group of unrecognised abusive head injury might have been prevented. Findings from these studies make a strong case that the initial scans must be interpreted by radiologists experienced in the field, or neuroradiologists, working closely with the clinician involved in managing the case. A standardised protocol for both scanning technique and interpretation will aid decisions at the district hospital.

The place for MRI versus CT or ultrasound scan are discussed in this paper. In virtually all (73/74) cases the first neuroimaging investigation was a CT scan, although one third of the cases thereafter had an MRI scan done within the next 12 days. While the CT scan better detects blood and bone, it is accepted that MRI is superior to CT in assessing brain parenchyma. Datta et al recommend that an MRI should be done in all cases. The rationale is that the MRI scan will better detect subdural haemorrhages in regions not easily visualised on the CT scan. In their whole group of NAHIs, 32/49 (65%) had subdural haematomas in more than one brain compartment, and they found an additional subdural by MRI scanning in 19 cases. These additional subdural haemorrhages, however, were not just confined to obscure brain areas. The overall frequency of subdural haematoma sites is more reliable from this large NAHI/imaging study than earlier ones, and was firstly, convexity, followed by interhemispheric fissure, posterior fossa, and then middle fossa.

A potentially controversial recommendation of Datta et al is that a CT scan should be strongly considered to be part of a skeletal survey for physical abuse, and in children under 6 months of age being investigated for possible child abuse. It is clear that neuroimaging is necessary in those with unexplained neurological symptoms and signs, and one has to ask what the likely benefits are of CT brain scanning for a child presenting with extracranial signs and symptoms of abuse, given the radiation dose. The potential benefit must depend on the chance that the latest episode of non-cranial abuse could have been predated by other inflicted injuries which did include brain injury. Evidence of repeated head injury (subdurals of different ages, etc) is reported to occur in 45% of children at first admission with suspected NAHI;⁸ for example, the child could have previously sustained an assault but not have been admitted to hospital, because of subtle symptomatology with a chronic, isolated, or slowly advancing subdural haematoma. If a positive CT brain scan signals the need for a skeletal survey (as is common practice), then a positive skeletal survey (particularly rib and metaphyseal fractures) should be followed by a CT brain scan.

There is a groundswell of consensus concerning the recommendations for imaging in non-accidental head injury cases. Inferring from the recommendations of Datta and colleagues,⁵ Jaspan and colleagues,⁶ Stoodley,⁹ and the British Society for Paediatric Radiology collectively, children with suspected NAHI (and children having a skeletal survey and skull films for suspected physical abuse, and infants under 6 months of age with sufficient grounds for considering the possibility of child abuse) should have an acute phase (day of presentation) brain CT scan to include brain and bone windows, contrast enhancement, and for children under the age of 2 years, brain slices of 3 mm in the posterior fossa and 5 mm elsewhere. A high resolution cranial ultrasound is undertaken on day 1–2.

If the original CT brain scan was abnormal, a brain MRI is performed on day 3 (or if not possible, a CT scan is repeated). The brain imaging should include the cervical spine and preferably the whole spine.

Follow up MRI is repeated at day 10–14, and in our practice, generally at 2 months, 6 months, and 2 years (unless otherwise clinically indicated) to follow the resolution of the acute injury and the evolution of the long term brain damage. Given the additional information provided by the MRI it can be reasoned that all such cases should have this imaging modality performed.